Will there be a 32b and 70b too?
#1
by
AlgorithmicKing
- opened
really appreciate the models but will there be a 32b and 70b too?
Thank you! Not planned, that would require using the original R1 model which needs much more compute and we don't have access to that kind of hardware unfortunately.
Also, the R1 tokenizer is a bit different even-though it's based on Llama, so it would require some work to figure out how to align the tokenizers otherwise we can't use the logits directly.